Rheumatology Advance Access originally published online on December 20, 2005
Rheumatology 2006 45(2):126-128; doi:10.1093/rheumatology/kei265
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
EDITORIAL |
The use of general practice consultation databases in rheumatology research
Primary Care Sciences Research Centre, Keele University, Keele, UK
Correspondence to: K Jordan. E-mail: k.p.jordan{at}cphc.keele.ac.uk
Clinicians, health service planners and public health policy-makers all understand the value of data which describe the occurrence of disease in the community. Routinely collected mortality statistics, for example, have made an enormous contribution to understanding the causes of disease, measuring the impact of prevention and health-care, and the planning of health services. However, a modern rheumatological readership will understand the point made forcibly in the 1980 Black Report on Inequalities in Health, that ‘undue dependence on mortality rates can induce comparative indifference towards problems of chronic illness’, since such indifference is still evident in health policy and research priorities today. The report emphasized the need for statistics on morbidity, especially chronic disease. The problem with morbidity is that, unlike death, much of the everyday experience of illness is not routinely recorded, whilst periodic surveys of the health of the general population, such as those carried out by the Office of National Statistics, whilst important, only provide a snapshot of the occurrence of selected problems at certain times.
One source of routine and continuing data on morbidity in the UK is general practice. More than 95% of people are registered with a general practice and the patient records held by those practices provide a picture of morbidity from cradle to grave. Furthermore, general practice is the first point of access to the National Health Service for most non-emergency care in the UK. Although many people with musculoskeletal symptoms seek advice and care from other sources, such as friends and neighbours, pharmacists, complementary therapists or private practitioners, most also use general practice. Most general practices in Britain are now recording at least some data electronically. This means that information on events arising from consultations in this setting—such as diagnosis, drugs prescribed, tests and referrals, as well as lifestyle data such as smoking status and body mass index—are becoming more readily available. Whilst this brings benefits for the organization and monitoring of clinical practice, such consultation data also provide an important source of information on the occurrence of symptoms and disease, and their treatment and care, and an important resource for researchers.
The growing importance of general practice databases is evidenced by the number of research articles generated from one of the largest samples of archived consultations in the UK, the General Practice Research Database (GPRD). Since 1990, 450 papers have been published using the GPRD. However, most have been about pharmacological issues, and the number of musculoskeletal research papers using GPRD or other similar databases is small. Researchers in rheumatology are, however, beginning to exploit the information which can be yielded by such consultation archives. The paper by Linsell and colleagues on a prospective study of patients consulting with shoulder problems using the Mediplus database [1] and a study by the same authors on patients with hip and knee problems [2] are examples of this.
GPRD and Mediplus are just two of an increasing number of large anonymized UK sets of routinely collected consultation data potentially available to researchers. For example, the GPRD contains data from around 300 practices across the country, covering an equivalent of 5% of the UK population. Others include the Weekly Returns Service run by the Royal College of General Practitioner's Birmingham Research Unit [3], QRESEARCH [4] and the Doctor's Independent Network (DIN) [5]. There are also regionally organized databases, such as our local North Staffordshire Consultations in Primary Care Archive (CiPCA), which contains validated data from 10 local practices [6].
General practice data routinely recorded about every contact offer the attractive possibility of charting the full range of musculoskeletal morbidity and how it is managed over time. There are caveats to this optimism, however, and as these databases become more widely available the positive reasons to use them must be considered in the context of potential constraints on their quality and interpretation. The strengths and weaknesses of specific databases have been discussed previously [7], but researchers need to be aware of aspects of general practice consultation recording that may be considered as limiting its value for research.
The first and most obvious point is that many general practices may not yet be recording all patient contacts on their computerized systems or systematically coding the reasons for those contacts. Various methods have been used to assess the quality of morbidity recording in general practice [8]. For example, Hassey et al. cross-referenced diagnostic codes with relevant drug prescriptions and a recorded seropositive rheumatoid factor to check on the completeness of records of rheumatoid arthritis in one practice [9]. The quality of prescription recording has tended to be better than that for recording morbidity [10]. However, the situation is improving as general practitioners (GPs) become more familiar with electronic recording. Computerized morbidity data are now being used by the government to assess the quality of care being provided and to renumerate practices on that basis. This should encourage both the completeness of computerized recording of patient contacts and the accuracy of coding. One of the claims of some of the large systematically organized databases is that they only use data from practices signed up to training and rules about regular recording of contacts. However, each database follows rather different rules which may not necessarily reflect what the average GP does. In the GPRD, for example, GPs only have to code morbidities on the first occasion and when a treatment is first issued. Whilst the GPRD appears an excellent resource for pharmacology research, these instructions mean that its use in exploring patterns of consultation may be more limited than that of other databases.
The second issue—and one that creates more doubts for the specialist rheumatologist than the first—is that there are no standardized methods of applying diagnostic labels in general practice. GPs are faced with an array of alternative codes for the same or similar conditions. One of the advantages of the coding systems used in practice is that the label applied can be a presenting symptom (‘shoulder pain’), but a disadvantage is that the basis for applying a diagnostic label, such as subacromial bursitis, may vary from GP to GP. The decision to use a label of joint pain rather than osteoarthritis may be down to habit or to whether the GP has decided to refer the patient for X-ray. Diagnostic criteria, however, are fluid in all clinical settings, as rheumatologists will appreciate, and consistency of diagnostic labelling is often no more a demonstrably strong feature of specialist practice than it is in general practice. As long as consultation data such as those in the Linsell paper [1] are interpreted as what the GP regards as shoulder pain, bursitis or capsulitis, then much of value can be taken from data collected in this way.
The third issue is that general practice consultation data is only providing a measure of morbidity for which people seek health care. However, there are advantages to this. First, consultation occurrence can be calculated as population rates because each practice has its specified registered population pool from which all consulters come. Second, consultation rates, although underestimating morbidity, are likely to reflect underlying patterns of occurrence of illness in the population, and the more severe and precise the diagnostic syndrome, the more likely this is to be true.
Finally, and well illustrated by Linsell et al. [1], general practice data provide one direct measure of health-care use. There are few publications, for example, on the proportion of musculoskeletal patients referred over time from general practice to specialist care, and it is a merit of the Linsell paper that it tackles this. The same concerns about validity apply: are all referrals recorded and are other routes to the specialist (casualty, private clinics) eventually entered in the practice records? Of course, many other routes of referral, such as therapies, may be less well recorded, and few of the many sources of health-care that lie outside the National Health Service (NHS) have established methods of routine recording. It is an incomplete picture, but a picture worth having.
The most obvious application of general practice consultation data to musculoskeletal morbidity is in determining the prevalence of symptoms and disease based on persons consulting and on the general practitioner's diagnostic label. Estimates of consultation prevalence for a number of musculoskeletal diseases, such as those provided by the Arthritis Research Campaign [11], are based on the most recent of four national surveys of general practice consultations carried out in 1991/92 (MSGP4 [12]). Such national practice samples are particularly valuable for providing estimates of the prevalence of rarer diseases.
Annual prevalence rates are usually defined as the number of people who have consulted about the problem at least once in the year, expressed as a proportion of all the population registered with the practices in the database. These will only pick up patients who consulted with that problem and were given a code for it during that year. Those who have the disease but have not been diagnosed and all others who did not consult for that problem during the year will remain invisible. An alternative is to measure prevalence over a longer period of time by expanding the search to cover several years and include those who only intermittently seek or use primary health-care. However, annual prevalence rates are also useful inasmuch as they provide a picture of current use of primary care for a particular problem.
Consultation databases would seem particularly suited to the measurement of incidence, given that the first onset of a problem is likely to be seen and recorded in primary care. A challenge is how to identify a consultation as being ‘first ever’, given that many patients do not have full coded histories on computer databases which may only have been running for a few years. Even if a complete history were available, most musculoskeletal conditions do not have a clear ‘first ever’ onset, and for problems such as back or joint pain it is the onset of a new episode or a recurrence that is likely to be most easily measurable. Use of episode coding (such as the system used to define incident cases in the MSGP4 into ‘first’, ‘new’ or ‘ongoing’) relies on the accuracy of the information obtained and entered by the health-care professional. The alternative is to work on the basis that a patient must have a certain number of years free from consultation for that problem prior to their current consultation. For example, Linsell et al. in their paper in this issue of Rheumatology and their previous publication [1, 2] use a 3-yr time period without a record of consultation for the morbidity to define incident cases of shoulder problems and hip and knee problems. The required number of years may vary from 1 for acute or chronic conditions requiring regular GP services to several for chronic conditions that can be managed by self-care or in secondary care.
General practice consultation databases can be used for the longitudinal study of disease. Time trends in the prevalence and incidence of disease have been studied in, for example, diabetes [13]. However, caution is needed as to whether these represent real changes of prevalence or improvements in morbidity recording or changes in diagnostic criteria. More ambitiously, individuals can be followed over time. Registered practice populations are dynamic rather than static as people move in and out of the practice, and, whilst this must be taken into account in following up cohorts of patients within general practice databases, it does not devalue such work. Potential longitudinal analyses include studies of long-term outcomes related to treatment, such as the risk of myocardial infarction in patients taking cyclooxygenase 2 inhibitors and other NSAIDs [14]. It may be possible to follow patients from before the first record of a condition to assess the prior history and predictors of diagnosis and consultation for conditions such as knee osteoarthritis [15]. After diagnosis, treatment and referral patterns can be investigated longitudinally (as Linsell et al. have done with shoulder conditions and hip and knee problems), as can comorbidities [16] and clinical outcomes.
Another advantage of population-based consultation data is that they can be linked to other characteristics, such as deprivation. Variation in the prevalence of consultation for rheumatoid disorders in New Zealand by deprivation status has been found [17]. Currently, the linkage of patients to deprivation scores within UK databases is limited, although this has been done in CiPCA and QRESEARCH. Deprivation, for example, has been shown to be associated with lower achievement of quality indicators for diabetes in the QRESEARCH database [18].
There is much potential in the creative and careful use of consultation databases for musculoskeletal research. Indeed, if such databases are routinely anonymized and downloaded as part of quality control and data collection in clinical general practice, they might in the future provide the most enduring measure of morbidity ascertainment, especially as survey research becomes increasingly constrained by ethical considerations and declining response rates. Nurses and therapists and other primary care team members are actively contributing their own coded data to computerized systems, and the recording of secondary care information, such as letters from rheumatology units, is improving, This means that the databases may soon provide a relatively complete picture of NHS health-care use for musculoskeletal disorders. Although there is still some way to go before this is a reality, continuing improvements in the quality of morbidity coding and the amount of information collected mean that general practice consultation databases will remain a rich resource for investigating the occurrence and course of musculoskeletal illness and disease in the community.
We are engaged in joint work on the use of general practice consultation data sets for musculoskeletal conditions with colleagues at the ARC Epidemiology Unit in the University of Manchester (Professor Deborah Symmons, Ms Alexandra Clarke), at the RCGP Birmingham Research Unit (Dr Douglas Fleming) and with the CiPCA team at Keele University. The views in the article are our own, but have benefited from discussion with these colleagues.
Independent grants from Medical Research Council and NCCRCD have allowed us to investigate the databases concerned (GPRD, RCGP Weekly Returns Service, CiPCA).
The authors have declared no conflicts of interest.
References
- Linsell L, Dawson J, Zondervan K et al. Prevalence and incidence of adults consulting for shoulder conditions in UK primary care; patterns of diagnosis and referral. Rheumatology 2005;45:215–21.[Medline]
- Linsell L, Dawson J, Zondervan K et al. Prospective study of elderly people comparing treatments following first primary care consultation for a symptomatic hip or knee. Fam Pract 2005;22:118–25.
[Free Full Text] - RCGP Birmingham Research Unit. Weekly Returns Service Annual Prevalence Report 2001–2003. London: Royal College of General Practitioners, 2004.
- Hippisley-Cox J, Stables D, Pringle M. QRESEARCH: a new general practice database for research. Inform Prim Care 2004;12:49–50.[Medline]
- Carey IM, Cook DG, De Wilde S et al. Developing a large electronic primary care database (Doctors’ Independent Network) for research. Int J Med Inform 2004;73:443–53.[CrossRef][Web of Science][Medline]
- Porcheret M, Hughes R, Evans D et al. Data quality of general practice electronic health records: the impact of a program of assessments, feedback, and training. J Am Med Inform Assoc 2004;11:78–86.[CrossRef][Web of Science][Medline]
- Majeed A. Sources, uses, strengths and limitations of data collected in primary care in England. Health Stat Q 2004;21:5–14.[Medline]
- Jordan K, Porcheret M, Croft P. Quality of morbidity coding in general practice computerized medical records: a systematic review. Fam Pract 2004;21:396–412.
[Abstract/Free Full Text] - Hassey A, Gerrett D, Wilson A. A survey of validity and utility of electronic patient generated records in a general practice. BMJ 2001;322:1401–5.
[Abstract/Free Full Text] - Thiru K, Hassey A, Sullivan F. Systematic review of scope and quality of electronic patient record data in primary care. BMJ 2003;326:1070–2.
[Abstract/Free Full Text] - Symmons D, Asten P, McNally R, Webb R. Healthcare needs assessment for musculoskeletal diseases: the first step – estimating the number of incident and prevalent cases. Chesterfield: Arthritis Research Campaign, 2002.
- McCormick A, Fleming D, Charlton J. Morbidity statistics from general practice: Fourth national study 1991–1992. London: HMSO, 1995.
- de Lusignan S, Sismanidis C, Carey IM, DeWilde S, Richards N, Cook DG. Trends in the prevalence and management of diagnosed type 2 diabetes 1994–2001 in England and Wales. BMC Fam Pract 2005;6:13.[CrossRef][Medline]
- Hippisley-Cox J, Coupland C. Risk of myocardial infarction in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis. BMJ 2005;330:1366–9.
[Abstract/Free Full Text] - Bedson J, Jordan K, Croft P. The prevalence and history of knee osteoarthritis in general practice: a case-control study. Fam Pract 2005;22:103–8.
[Free Full Text] - Kadam UT, Jordan K, Croft PR. Clinical comorbidity in patients with osteoarthritis: a case-control study of general practice consulters in England and Wales. Ann Rheum Dis 2004;63:408–14.
[Abstract/Free Full Text] - Taylor W, Smeets L, Hall J, McPherson K. The burden of rheumatic disorders in general practice: consultation rates for rheumatic disease and the relationship to age, ethnicity, and small-area deprivation. N Z Med J 2004;117:U1098.[Medline]
- Hippisley-Cox J, O’Hanlon S, Coupland C. Association of deprivation, ethnicity, and sex with quality indicators for diabetes: population based survey of 53,000 patients in primary care. BMJ 2004;329:1267–9.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Barber, S. Muller, T. Whitehurst, and E. Hay Measuring morbidity: self-report or health care records? Fam. Pract., February 1, 2010; 27(1): 25 - 30. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Kush, L. Alschuler, R. Ruggeri, S. Cassells, N. Gupta, L. Bain, K. Claise, M. Shah, and M. Nahm Implementing Single Source: The STARBRITE Proof-of-Concept Study JAMIA, September 1, 2007; 14(5): 662 - 673. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||